goodroot/hyprwhspr

Native speech-to-text for Linux - Fast, accurate, private and hackable system-wide dictation

instant performance | Cohere / Parakeet / Whisper / Qwen3-ASR / ElevenLabs / REST API | stylish visuals

Supports Arch, Debian, Ubuntu, Fedora, openSUSE and more

hyprwhspr.com

• Built for Linux - Native AUR package for Arch, or use Debian/Ubuntu/Fedora/openSUSE/NixOS
• Very fast defaults - Nearly instant and accurate performance via in-memory models
• Offline and private - Transcription never leaves your machine. Cloud APIs are fully optional.
• Latest models - Cohere Transcribe? Orukeet? Parakeet TDT V3? Qwen3-ASR? Latest and greatest
• GPU memory efficient - Limit or zero memory usage easily, more for other local models
• onnx-asr for wild CPU speeds - No GPU? Optimized for great speed on any hardware
• Strong CJK - Qwen3-ASR for Chinese, Japanese and Korean
• Translation - Translate non-English to English with a single config
• REST API or websockets - Secure, fast wires to top clouds like Gemini, ElevenLabs
• Live streaming - Self-host NeMo or Phonon: watch words form, or type as you speak (experimental)
• Themed visualizer - Visualizes your voice, will automatch Omarchy and Noctalia themes
• Word overrides and prompts - Custom hot keys, common words, and more
• Multi-lingual - Great performance in many languages
• Long form mode with saving - Pause, think, resume, pause: submit... Bam!
• Auto-paste anywhere - Instant paste into any active buffer, or even auto enter (optional)
• Audio ducking 🦆 - Reduces or pauses other audio on record (optional)

Why hyprwhspr? There are a lotta dictation apps. This one is built ground-up for performance and stability: your hardware will auto-optimize to the best possible accuracy and speed. Actively maintained, at the leading edge, fully featured, strictly offline unless you use an API, and works on both Wayland and X11.

• Linux with systemd (Arch, Debian, Ubuntu, Fedora, openSUSE, NixOS etc.)
• Python 3.11-3.14
• Wayland or X11 session (GNOME, KDE Plasma, Sway, Hyprland, Niri, etc.).
• Clipboard/window tools:
  • wl-clipboard and wtype on Wayland
  • python-pyperclip, xclip, xdotool, and xprop on X11 (installed by the dependency script)
• Waybar or Noctalia (optional, for status bar)
• gtk4 + PyCairo (optional, for visualizer)
• NVIDIA GPU (optional, for CUDA acceleration)
• AMD/Intel GPU / APU (optional, for Vulkan acceleration)

On the AUR:

# Install for stable
yay -S hyprwhspr

# Or install for bleeding edge
yay -S hyprwhspr-git

Then run the interactive setup:

hyprwhspr setup

curl -fsSL https://hyprwhspr.com/install.sh | bash

Installs dependencies, clones to ~/.local/share/hyprwhspr/src, and walks you through setup.

# Clone the repo
git clone https://github.com/goodroot/hyprwhspr.git
cd hyprwhspr

# Install dependencies for your distro
./scripts/install-deps.sh

# Run interactive setup
./bin/hyprwhspr setup

Use the Nix flake and NixOS module for x86_64 Linux with Whisper.cpp CPU transcription.

We'd appreciate issue reports and requests!

Setup then walks you through:

1. ✅ Configure transcription backend (Cohere Transcribe, Parakeet TDT V3, Whisper, Qwen3-ASR, REST API, or Realtime WebSocket)
2. ✅ Download models
3. ✅ Configure themed visualizer for maximum coolness (optional)
4. ✅ Configure bar integration for your shell -- Waybar or Noctalia (optional)
5. ✅ Set up systemd user services
6. ✅ Set up permissions
7. ✅ Validate installation

Ensure your microphone of choice is available in audio settings!

1. Log out and back in (for group permissions)
2. Press Super+Alt+D to start dictation - beep!
3. Speak naturally
4. Press Super+Alt+D again to stop dictation - boop!
5. Bam! Text appears in active buffer!

What you'll see while recording: on layer-shell compositors (Hyprland, Sway, niri, KDE) the animated mic OSD overlay -- on Noctalia / Omarchy it auto-matches your live shell theme; on GNOME/Mutter you may need to make additional changes. See Themed visualizer for details.

Any snags, please create an issue.

# Arch: update via your AUR helper
yay -Syu hyprwhspr

# Managed release installs
hyprwhspr update

# Legacy bootstrap installs: re-run the installer
curl -fsSL https://hyprwhspr.com/install.sh | bash

# Either way, setup is idempotent if you need to re-run it
hyprwhspr setup

Managed release rollout and recovery: installation guide.

• hyprwhspr setup - Interactive initial setup
• hyprwhspr update [--version VERSION] - Update a managed release
• hyprwhspr install status - Inspect managed installation state
• hyprwhspr install repair [--python PATH] - Repair a managed installation
• hyprwhspr config - Manage configuration (show / show --all / edit / secondary-shortcut)
• hyprwhspr model - Manage models (download / list / status / unload / reload)
• hyprwhspr record - External hotkey control (start / stop / toggle / cancel / capture / status / copy-last / paste-last / clear-last)
• hyprwhspr transcribe INPUT - Transcribe a WAV or MP3 file to stdout (use -o PATH to write a file)
• hyprwhspr status - Overall status check
• hyprwhspr validate - Validate installation
• hyprwhspr test - Test microphone and transcription end-to-end
• hyprwhspr keyboard - List/test keyboard devices (list / test)
• hyprwhspr waybar / hyprwhspr noctalia - Manage bar integration
• hyprwhspr mic-osd - Enable/disable the mic OSD (enable / disable / status)
• hyprwhspr systemd - Manage systemd services
• hyprwhspr uninstall - Remove hyprwhspr; preserve settings, credentials and models
• hyprwhspr uninstall --purge - Also remove recorded personal files (--keep-models preserves models)

All of it at hyprwhspr.com/docs, or here:

• Configuration -- the config file, validation, minimal setup
• Recording modes · Hotkeys · Paste · Text processing · Feedback
• Backends · Memory -- which model, and freeing it
• Troubleshooting

• Check logs: journalctl --user -u hyprwhspr.service
• Troubleshooting covers common issues
• Still stuck? Create an issue - logs helpful!

MIT License - see LICENSE file.

Start with an issue - happy to help! Pull requests best begin there too.

Upvotes on the AUR and/or GitHub stars are also a nice way to show appreciation.

Built with ❤️ in 🇨🇦